Ultrasonic Detection of Interlayer Cracks in P91 Steel Pipe Butt Welds
The Problem That Keeps Power Plant Welding Supervisors Awake
P91 is a 9Cr-1Mo martensitic steel developed for high-temperature service in power generation. It is a superb material when welded correctly, and a nightmare when it is not. The weld metal and the heat-affected zone (HAZ) are susceptible to a range of defects, and interlayer microcracks are among the most insidious. These cracks are small, often less than a few millimetres in length and height, and they sit between weld passes where the thermal history is complex. If you rely on radiography alone, you will miss them. If you rely on conventional ultrasonic testing (UT) with a standard acceptance criterion, you will either miss them or reject sound welds.
This paper by Wang Zhiyong and colleagues at the Hebei Electric Power Research Institute addresses exactly this problem. They present a case study of interlayer cracks in P91 butt welds, describe the ultrasonic response, and propose a combined testing strategy that includes radiography, UT, and time-of-flight diffraction (TOFD) for verification.
Why P91 Interlayer Cracks Are So Difficult to Detect
The difficulty has three roots. First, the cracks are small. They typically have a height of 1–3 mm and a length of 5–20 mm. Second, they are oriented roughly parallel to the weld surface, so the ultrasonic beam from a conventional angle probe strikes them at an unfavourable angle, producing a weak reflection. Third, the weld itself has a coarse-grained, columnar microstructure that scatters ultrasound, creating a high noise floor that masks the crack signal.
The authors report that the reflected wave height from these cracks is often only 20–30% of the reference level, which is below the typical recording threshold of 20% or the rejection threshold of 50% in many standards. A crack that is real but produces a 25% indication is easy to dismiss as a non-relevant indication, especially if the weld also contains the normal run-on and run-off signals.
The table below summarises the detection challenge.
| Detection Method | Crack Length ≥ 5 mm | Crack Length < 5 mm | False Call Rate | Remarks |
|---|---|---|---|---|
| Radiography (RT) | Reliable for through-wall defects | Poor for planar cracks | Low | Misses tight, thin cracks |
| Conventional UT (single probe) | Variable, depends on angle | Poor | High | Weak reflection, high noise |
| TOFD | Reliable for vertical extent | Reliable for small cracks | Low | Best for through-wall sizing |
| Phased Array UT (PAUT) | Good | Moderate | Moderate | Requires skilled operator |
The Case Study and the Proposed Strategy
The authors describe a specific case where a P91 weld was tested by RT and passed, but a subsequent UT examination revealed a suspicious indication. The indication was small, and the initial assessment was that it was a non-relevant reflector. However, the weld was later sectioned, and the section revealed a genuine interlayer crack.
The key lesson is that the UT indication, although small, had a characteristic that should have raised a flag: it was located at the boundary between two weld passes, and it had a high signal-to-noise ratio when the probe was moved to a specific position. The authors recommend that any such indication be investigated further using TOFD, which can size the crack height accurately and distinguish it from a planar reflector such as lack of fusion.
The proposed strategy is straightforward and, in my view, should be adopted as a standard practice for P91 welds:
- Perform RT as the primary volumetric examination.
- Perform UT in parallel, using a sensitivity that is at least 6 dB more sensitive than the standard reference level.
- For any UT indication that is above the recording level, or that has a suspicious location and orientation, perform TOFD to verify and size the defect.
- If TOFD confirms a crack, evaluate it against the acceptance criteria for planar defects, which are typically more stringent than for volumetric defects.
Practical Lessons from the Power Plant Floor
I have been involved in P91 welding for power plant piping for over two decades, and I have seen the consequences of inadequate NDT. The most common failure mode is not a catastrophic rupture during hydrotest, but a slow, insidious crack growth that leads to a leak after months or years of service. The crack starts as an interlayer microcrack, grows under creep-fatigue loading, and eventually breaches the wall.
My own experience aligns with the authors' recommendation to use UT in parallel with RT. I would add the following practical rules.
- P91 welds should never be accepted on the basis of RT alone. The probability of detecting a tight, planar interlayer crack with RT is unacceptably low.
- UT should be performed with a focused probe and a scanning plan that covers the full weld volume, including the interpass boundaries.
- The UT operator must be experienced with P91 welds and must understand the microstructure. A novice operator will dismiss the crack signal as noise.
- TOFD is the definitive method for sizing interlayer cracks. If TOFD is not available, a phased array UT with a sectorial scan can provide similar information.
- Any indication that is located at an interpass boundary and has a low amplitude but a consistent, repeatable response should be treated as a suspect defect until proven otherwise.
Standards Context and Remaining Gaps
The paper references the Chinese standard for ultrasonic testing of welds, which sets a recording level and a rejection level. The authors argue, correctly, that the standard's acceptance criteria are not appropriate for P91 welds because the crack signal is inherently weak. They propose a more sensitive approach, which is consistent with the philosophy of the ASME Boiler and Pressure Vessel Code, Section V, and with the European standard EN ISO 17640.
From a procurement and delivery perspective, I would add that the NDT strategy should be specified in the welding procedure qualification and in the inspection and test plan (ITP). If the client's ITP only requires RT, I would raise a technical query and request that UT be added. The cost of UT is small compared to the cost of a weld failure.
Study Insights and Engineering Judgement
This is a valuable paper because it addresses a real problem with a real case study and a practical solution. The authors have not tried to solve the metallurgical problem of why interlayer cracks form; they have focused on the detection problem, which is where the immediate risk lies. Their recommendation to use RT and UT in parallel, with TOFD as a verification tool, is sound and should be adopted as a best practice.
The one area where I would add a note of caution is the reliance on TOFD for acceptance. TOFD is excellent at sizing cracks, but it does not always provide a clear indication of the crack's nature. A lack of fusion can produce a TOFD response similar to a crack. Therefore, the final acceptance decision should be based on the combined evidence from RT, UT, and TOFD, and, if necessary, on a metallographic replica or a sectioning test.
Reference Value and Outlook
For a welding engineer or an NDT supervisor on a power plant project, this paper is a concise and practical reference. It provides a clear rationale for a combined NDT strategy and a specific recommendation for handling suspect indications. For a pipe manufacturer, it is a reminder that the NDT requirements for P91 welds are more demanding than for carbon steel, and that the inspection plan must reflect the material's unique behaviour.
The authors have highlighted a gap in the existing standards and have proposed a sensible solution. The next step should be a broader study that quantifies the probability of detection (POD) of interlayer cracks for different UT techniques, and that establishes a reliable acceptance criterion based on crack size rather than on signal amplitude. Until that work is done, the combined RT + UT + TOFD strategy should be treated as a mandatory requirement for P91 welds.
Zhuojin Pipe Fitting Co., Ltd